|
Track: New Technology and Technology Integration
Kan He
Missouri Resource Assessment Partnership
4200 New Haven Road
Columbia, MO 65201
Telephone: 573-875-5399
Fax: 573-876-1896
E-mail: kanhe@msc.nbs.gov
John Jensen
Urban Road Structure Guided Remotely Sensed Data Classification
Defining Issue: Automated urban area land use classification of satellite remotely sensed data is a difficult task due to the incompatibility between earth's natural spectral patterns and human-defined information categories. GIS Solution: Spatial information contained in GIS can play an important role in resolving the incompatibility between the spectral patterns and human-defined information categories. Natural or human-made linear features are often boundaries for homogeneous land information classes (e.g., a road or creek). Roads are built to improve spatial interaction between people. Therefore, the spatial pattern of road infrastructure will appear differently on different land use zones. A knowledge-based analysis approach was used in combining remotely sensed data with urban road structure to improve the accuracy of urban land use identification. Methodology: Road network information was used to divide the research area into simplified Road Bounded Regions (RBR). The geometrical
characteristics and the statistics about the remotely sensed spectral contents were evaluated for each RBR. Considering the result of the evaluation, several parameters were added to each RBR. A subsequent knowledge-based analysis and postspectral classification using these parameters resulted in a new classification map that agrees with the cartographic standard more closely than traditional remotely sensed image classifications.
|
|
